選擇卷期
- 期刊
- Ahead-of-Print
近年以「沈浸式」為標題的影片在抖音為主的短影音平台上激增，本研究探討短影音廣告中的廣告類型、拍攝視角、廣告字幕，如何藉由影響觀眾的沉浸體驗提高廣告的說服（廣告態度、品牌態度、購買意願）和滲入效果（品牌回想、辨識）。兩個實驗結果發現：功能型廣告搭配第一與第三人稱轉換拍攝視角，或體驗型廣告搭配第一人稱拍攝視角，可產生較強的沉浸體驗和廣告效果；體驗型、第一人稱拍攝視角之短影音，無字幕（vs.有字幕）產生較強的沉浸體驗和廣告效果；功能型、第三人稱拍攝視角之短影音，有字幕（vs.無字幕）不影響沉浸體驗，但可直接提升態度反應和滲入效果。本研究對短影音廣告、沉浸體驗具有理論貢獻，並為短影音廣告創作者提供實務建議。
- 期刊
- Ahead-of-Print
The successful integration of newcomers into online brand communities is essential for sustaining community vitality and cohesion. The integration of newcomers frequently introduces fresh perspectives and new social connections, affecting interaction patterns in online communities. This study addresses the critical challenge of understanding how newcomers adjust and how this adjustment impacts their participation. Grounded in socialization theory, we conceptualize newcomer adjustment as a comprehensive second-order construct comprising self-efficacy, role clarity, and social acceptance. By leveraging both self-reported and archival behavioral data, we demonstrate that effective newcomer adjustment is key to fostering active participation within online communities. Our findings reveal that community insiders' factors such as member similarity, involvement, and receptivity significantly facilitate newcomer adjustment. Moreover, this study explores the moderating role of pride-a self-conscious emotion-in strengthening the relationship between newcomer adjustment and participation behaviors. We find that pride enhances confidence, encourages generosity, and deepens social bonds, thereby amplifying the positive effects of newcomer adjustment on participation. By integrating both information sharing and virtual gift-giving into our comprehensive model, we offer a roadmap for enhancing newcomer integration and, ultimately, ensuring the long-term success and vitality of online brand communities.
- 期刊
- Ahead-of-Print
The study focuses on improving the accuracy of Gross Domestic Product (GDP) forecasting by identifying essential variables using feature selection techniques and machine learning algorithms. The study conducted experiments using four well-known feature selection techniques and five prediction methods and found that linear regression combined with single feature selection outperformed all combinations with the lowest error rate. The study identified 17 essential variables contributing to GDP prediction accuracy, highlighting that local industry is the most important factor in predicting Taiwan's GDP. The study also found that a variable selected by multiple feature selection methods simultaneously reflects its importance, not its predictive accuracy. In addition, to determine the performance of the models under a time series, we conducted the experiments, and the Random Forest algorithm produced the lowest error rate under the Lag-8Q strategy. The theoretical and practical implications of the study suggest that a smaller set of essential variables is sufficient to build a better GDP forecasting model instead of using many economic and financial indicators. The study's findings can contribute to developing a more efficient and effective GDP forecasting model that focuses on a smaller set of essential variables. In addition, policymakers and public officials can use the study's findings to monitor the country's macroeconomy more closely and make informed decisions for related policies. However, the study has potential limitations, such as the 17 most important variables identified may not be generalizable to other entities. For future work, the study suggests that new feature selection methods and combination strategies can be proposed to improve the accuracy of GDP prediction.
- 期刊
- Ahead-of-Print
Conspicuous travel experiences have been exhibited on social media, changing e-marketing strategy. However, most existing studies have investigated this topic based on a hindsight view. As decision-making (like making a conspicuous post) involves future-focused considerations, and social media usage is a kind of technology usage, the future-oriented perspective and the dual-factor model of technology usage are regarded as the theoretical foundation. Based on this foundation, the study examines the impacts of enablers (social status improvement and relationship maintenance) and inhibitors (anticipated guilt and anticipated social anxiety) on the intention of posting conspicuous travel experiences on Instagram. Data were gathered from 362 Instagram users through an online survey and analyzed using a structural equation model with SmartPLS software. The findings show social status improvement and relationship maintenance positively influence on the intention to post, whereas anticipated guilt and anticipated social anxiety have negative effects. The study deepens the current understanding of the relationship between future-oriented emotions and intention to post conspicuous travel experiences, and supplements the dual-factor model of technology usage by leveraging it to explain intention to post travel experiences.
- 期刊
- Ahead-of-Print
This study leverages Natural Language Processing (NLP) techniques to detect stances in Chinese online news articles. With the increasing consumption of news through the internet, media bias and confirmation bias have become more prevalent. This research aims to address these biases by classifying the stances of news articles on controversial issues such as diversified families, nuclear power generation, abolition of the death penalty, and legalization of abortion. Using a dataset of news articles collected from top Taiwanese news media over the past years, we employ BERT-based models to classify news stances as supportive, neutral, or opposing. Our findings indicate that using news content as input significantly improves the accuracy of stance classification. Moreover, the study demonstrates that removing neutral content enhances the model's performance. The results suggest that BERT-wwm is more effective for complex classification tasks, while SVM-Linear performs better for simpler classifications. In the data labeling stage, this study uses the method of labeling news headlines to find sentences related to the stance. In the SVM-Linear method, the model's performance reaches more than 80%. Based on this model, the auxiliary mark of news content is constructed. The model uses BERT as the primary model. The input part is adjusted to the training data set to improve the accuracy of news stance classification. This research contributes to the field of stance detection by providing a robust methodology for analyzing media bias and enhancing the accuracy of news stance classification for issues stance information disclosure.
- 期刊
- Ahead-of-Print
本研究旨在透過Landsat-8衛星遙測數據，建立美國玉米帶的玉米產量預測模型。採用高空間解析度的方法，成功克服傳統作物產量預測方法的限制，並有效捕捉玉米生長狀態和產量變異性。研究顯示，整合多種植被指數與作物分布面積，同時考慮物候管理的差異性，基於機器學習所建立之非線性迴歸模型能有效預測玉米產量，並提升模型泛化能力。本研究採用多個回歸模型，其中非線性支持向量迴歸模型R^2最高可達0.94，均方根誤差僅0.06。精準農業數據將促進期貨、金融機構及保險業等商業領域在電子商務中的應用，並有助於實現SDG 2的目標倡議。
- 期刊
- Ahead-of-Print
深度學習擁有優異的特徵學習能力，支持向量機擁有優異推理能力。近年來，完美結合兩者優點的深層支持向量機網路吸引眾多學者的關注，與傳統深層神經網路相比，深層支持向量機網路有下列優點：（1）具有較高的推理能力；（2）更適合在訓練樣本數目不足的任務。本論文提出一個嶄新的深層模糊對偶支持向量迴歸網路，透過股價的數值資料來預測股價的變動。本研究融合：（a）演化計算，（b）集成學習，（c）深度學習與（d）多核心函數學習的優點。本研究提出的深層模糊對偶支持向量迴歸機除了能提供最可能的預測結果，還可以提供預測結果的模糊範圍的內界與外界，以及提供預測結果的信心程度，這對於對於股票買賣的決策制定任務是很重要的。
- 期刊
- Ahead-of-Print
網路時代來臨，單純文字、圖片或影片，無法完整呈現傳達內容，直播技術成熟後，直播主與觀眾間可以更自由互動。本研究以類社會臨場感與社會認同理論為基礎架構，探討在網路直播的情境下，是否使觀眾更直接、深入地了解直播內容，進而減少觀眾在認知上的誤差，提高持續觀看意圖。本研究以觀看過網路直播者為受測對象，透過問卷調查方式收集樣本，以線性結構方程式檢驗本研究之架構與假說是否成立。研究結果顯示，類社會臨場感中的「理解感」、「正面觀」、「參與感」、「主宰感」能透過觀眾對網路直播主或群體之認同，進而影響持續觀看意圖。直播主規劃直播內容時，要從觀眾的角度思考，滿足需求並持續吸引他們的好奇與參與，以達到網路行銷目的。
- 期刊
- Ahead-of-Print
The aim of this research was to examine the determinants of mobile-banking technology in order further to provide an integrated model on which the acceptance of innovative services in relation to the behavioral intention consumers have to use food delivery platforms can be based. This cross-cultural study was conducted in Taiwan and Thailand. The results indicated that functional risk, security, and supply uncertainty could be critical determinants, with direct related innovative service risks. Also, based on the results, the innovative service risks could influence the intentions users have when they use the m-banking system. This is because immediate enhancement in their behavioral intention of using food delivery platforms can be made. Finally, the empirical findings could provide researchers with valuable theoretical contributions and practical implications in the food delivery platform sector.
- 期刊
- Ahead-of-Print
近年來，網紅代言成為重要的行銷方式。然而，目前尚缺乏能全面衡量網紅影響力的量表。本研究發展網紅影響力量表，透過四階段的實證達成此目標。首先，文獻探討確定網紅影響力構面，並發展初步量表。隨後階段，以探索性因素分析純化量表並驗證信效度，獲得45題完整語句版與45題簡要敍述版量表。第三階段比較45題簡要敍述版與本階段15題短版，最終確定15題短版量表的適用性。第四階段則利用多元特質多元方法矩陣（Multi-Trait-Multimethod Matrix, MTMM）驗證量表的聚斂性與區辨性。最終，本研究發展出包含四個維度與15個構面的網紅影響力量表，並提供兩個版本（45題簡要敍述版與15題短版）。此量表將為行銷工作者提供評估網紅影響力的指標，並為未來學術研究奠定基礎。